A range of whole seconds against a menu of three
One page states an integer in the range two to thirty seconds. The other states four, six or eight, with eight available only at 1080p, 4k or with reference images. A range and a conditional menu are not the same kind of fact, and a shot list cannot be written the same way against both. As of 2026-09-22.
| What the pages state | Wan 3.0 | Google Veo |
|---|---|---|
| Form of the offer | An integer range | A menu of three values |
| Ceiling | 30 seconds | 8 seconds |
| Condition | Input plus output must not exceed 30 | 8 seconds needs a tier or references |
Inclusion rule. What each page states about the length parameter, kept in the form the page uses. A menu is not rewritten as a range. Order. The form of the offer first, then the numbers inside it.
1A range lets the edit lead; a menu makes the edit follow
Where any whole value is available, the beat is written first and the request follows it. Where three values exist, the beat is written to one of them, and a seven second idea becomes a six second idea or an eight second idea before anything is generated.
That is a creative constraint arriving through a parameter, which is the kind this register was built to surface. Neither page frames it that way, and both are describing it accurately in their own terms.
2The conditions are attached to different things
One condition is about extension: supply a video and the input shares the thirty second allowance with the output. It binds only on the conditioned path. The other is about quality: the longest value requires a higher tier or reference images, and it binds on every request that wants eight seconds.
So the first condition is invisible to most jobs and severe on the ones it touches, and the second is unavoidable and mild. Averaging them into a single verdict about which page is more permissive would lose both facts.
3Join counts for the same slot
Thirty seconds is one request on the first row and four on the second. Sixty seconds is two and eight. Every one of those extra seams is a place where separately generated footage has to match, and nothing on either page speaks to whether it will.
The register stops at the count. It is arithmetic on published figures, and it is the number that decides whether continuity becomes the main creative problem of the job or never comes up.
4The rest of the column behind these two
| Generator | Duration |
|---|---|
| Amazon Nova Reel | 6 s increments, up to two minutes |
| CogVideoX | 6 s, at 49 frames |
| Google Veo | 4, 6 or 8 s |
| Grok Imagine | 1 to 15 s |
| Hedra | 10 minutes |
| HeyGen | 30 minutes per scene |
| Kling AI | Not stated |
| LTX-Video | Frames, never seconds |
| MiniMax | 4 to 15 s |
| Pika | 5 s and 10 s, 20-25 s on Pikaframes |
| PixVerse | 1 to 15 s on version 6 |
| Runway | 5 to 15 s on Gen-4.5 |
| SceneMixer | Thirty seconds on the Wan tiers, fifteen on H3 |
| Vidu | Not stated |
| Wan 3.0 | 2 to 30 s |
Inclusion rule. Every generator this register carries, on the one column this pair disagrees about. Rows that state nothing keep their place and are marked hollow. Order. Alphabetical by generator name.
Why these two. The ceilings differ by a factor of nearly four, and the offers differ in kind: one lets a beat be asked for at its real length, the other makes the beat fit one of three values and attaches a tier to the longest.
- Duration parameterAn integer in the range [2, 30] seconds, and input plus output must not exceed 30 seconds
- Veo 3.1 and Veo 3.1 Fast8 seconds, 6 seconds, 4 seconds, with 8 seconds only if 1080p or 4k or using reference images
5Sources
Read from the Wan3.0 API reference and Google's Veo documentation, both checked 2026-09-22. The column is described on duration, and each entry has its own cell page at Wan 3.0 and Google Veo, and the rule that keeps surfaces and generators apart is on method. Nearby pairs: Nova Reel and Runway, CogVideoX and Kling AI.